The hand, the most complex biological machine of mankind, is still the most challenging unsolved problem in the field of robotics. If engineers can overcome this hurdle, the robots being developed in labs today may become a common sight on factory floors in the future.
Robotics engineers around the world are working hard to solve a cutting-edge problem in the fields of artificial intelligence and mechanical engineering: how to build a robot hand that can operate like a human hand. At present, although humanoid robots can already walk, lift and balance, their lack of dexterous and perceptive hands is still one of the main obstacles to their large-scale application in factories and various workplaces.
Researchers say the goal is not just to make robots more human-like in appearance, but to equip them with the ability to perform the fine, complex movements required for most skilled labor. Tesla’s humanoid robot Optimus, as a representative, is directly challenging this problem. According to Morgan Stanley's estimates, if this threshold can be exceeded, the global humanoid robot market may be worth up to US$5 trillion by 2050.
“For robots to be truly useful, they have to have amazing hands,” Elon Musk said in an interview with the Wall Street Journal.
Although Optimus has been able to achieve bipedal walking, Musk has also said that designing human-like hands is much more difficult than walking itself.
A research team at Northwestern University's Center for Robotics and Biological Systems is designing a highly sensitive and flexible robotic hand through a federally funded project. Team leader Kevin Lynch pointed out that they have set a ten-year goal to give robots the flexibility to complete basic human tasks.
Take the prototype in Lynch's lab, for example. The hand is based on a model from British company Shadow Robot. Several cylindrical motors the size of coffee cans drive mechanical fingers. Sensors installed on the fingertips can sense changes in the electrical properties of liquids similar to those under the "skin". When a finger touches an object, the sensor converts these changes into "tactile"-like data in real time.
Graduate students continue to train hand coordination by arranging simple links for the robot - such as looping, holding blocks, and guiding small objects. The collected data is used to improve the performance of machine learning algorithms. Lynch said that in order to complete fine movements such as "writing with a pencil", future versions will need to add more sensors to the fingertips and palms.
In addition, there are also research teams that are breaking through the shackles of "humanoid form". The four-finger manipulator developed by Matei Ciocarlie, a professor of mechanical engineering at Columbia University, can judge the shape and material of objects by touch alone, making up for the lack of vision. This hand can lift fragile items, such as paper tubes, but may occasionally slip or fall.
Boston Dynamics is taking a different route. Its experimental humanoid robot Atlas is equipped with three-fingered hands that can be flexibly transformed into a "thumb-type grip" or a "paddle-shaped palm". The video it released shows that Atlas can lift car accessories, balance dumbbells, and grasp small objects. Project leader Alberto Rodriguez said that this design always seeks a balance between strength, dexterity, slenderness and durability. "It is not enough to design a weak and inefficient gripper alone."
It’s worth noting that not all engineers are pursuing “human-like hands.” Igor Kulakov, co-founder and CEO of MicroFactory from San Francisco, prefers to simplify industrial design. Their $5,000 robot has a dual-arm design, with one hand holding specialized tools and the other using two fingers to hold objects. This configuration can complete key manufacturing steps such as welding circuit boards, tightening screws, and removing protective films, and the cost is much lower than that of complex humanoid robots.
Despite some progress, materials science still faces stubborn challenges. Rich Walker, director of Shadow Robot, said that current manufacturing processes are still difficult to replicate the basic functions of the human body, such as self-healing skin and self-lubricating joints, which pose many obstacles to product engineering.
The drive to imitate human hands partly stems from the increasingly severe "labor shortage" in the manufacturing and nursing industries. Ed Colgate, a professor of mechanical engineering at Northwestern University, pointed out that improving the dexterity of robots may make automation tools affordable for small and medium-sized enterprises, and they will no longer be the exclusive preserve of large enterprises. "It could also create entirely new job opportunities. That's what motivated us to do this research."